• Print

How the Zeo sleep device works around the limitations of home monitoring

The
Zeo
is part of a trend toward using technology to monitor our own bodies.
People have always been concerned about their health, of course, and
have tried different things to see what works (including rather absurd
superstitions). But now there are ways to bolster one’s curiosity with
real scientific data.

ZEO_HAND_HB_rgb.jpg

The Zeo makes this data available for people
who may have sleep problems–and quite a lot do, judging from a

2005 National Sleep Foundation (NSF) poll

(a recent poll covering a wide range of adults):

26% say they had “a good night’s sleep” only a few nights a month or
less. Another 24% report having “a good night’s sleep” a few nights a
week.

This blog is meant to be a technical discussion, not a consumer guide,
so I took the opportunity to talk to Ben Rubin, CTO and cofounder of Zeo, just a couple weeks
after the official release of the Zeo to get some information on two
aspects:

  • How it collects data during sleep

  • How they analyze the data to help the customer sleep better

The Zeo sensor device is a simple headband with a silver layer that
lies on your forehead. The designers chose silver because it doesn’t cause
allergic reactions and efficiently picks up electronic waves with
minimal noise from sweat or other interference.

The electrical signals given off by the brain, the eyes, and muscle
movements around the face are collected 128 times per second.
Activity at different frequencies indicates what kind of sleep you’re
in (if any), in ways well known to sleep researchers. For instance:

  • Activity in the 11 to 14 Hz range indicate some form of sleep. Spikes
    in that range–sleep spindles–indicate light sleep.

  • Activity in the 2-4 Hz range (Delta waves) indicate deep sleep.

  • REM sleep has less in the Delta range and more in higher frequencies,
    along with the heightened eye movements that give the sleep phase its
    name, and fewer muscle contractions, which can be measured by activity
    in the 30 Hz range.

And so forth. The headband knows the essential frequencies of
electronic activity in the brain, in the eyes (because they’re
electronically polarized) and in the facial muscles. This is quite
impressive when you look at traditional sleep labs. They put sensors
on many different parts of the face to get their data. In contrast,
the Zeo picks up a signal that it has to separate into frequency
bands. Of course, the lab is more accurate, but its cost makes it
feasible only for diagnosed sleep disorders. (The Zeo company wants to
make it clear that they aren’t providing a medical device and don’t
offer diagnoses.)

Samples are run through a Fast Fourier Transform and more complex,
proprietary algorithms every 30 seconds to get data that the Zeo
device on your bedstand accumulates. It can tell you later how many
hours you spent in each phase of sleep. Every five minutes, another
calculation determines what phase you’re in (light, REM, deep, or
awake) and displays a band on an LED screen so you can check the
pattern of your sleep when you wake up. (In the figure, the bands indicating heaviness of sleep appear near the bottom of the screen.)

(Update, December 22: Zeo developers have posted

more detail about the timing of data collection
.)

ZEO_FRONT_8 32_ZQ73.jpg

The calculations used by the Zeo were developed through a neural
network trained on huge numbers of people, both normal sleepers and
troubled sleepers. In effect, the algorithms were refined by comparing
their output to the analysis of sleep experts for each person who
joined the experiment. (Rubin was one of the guinea pigs, along with
many other staff. That was probably the most sleep any of them got
during their work in this start-up.)

Zeo ran an experiment in which sleepers were simultaneously monitored
by a prototype of the Zeo, by an older technology called an actigraph,
and by sleep researchers using a polysomnograph. The calculations by
the Zeo prototype came fairly close to the accuracy of the other forms
of measurement. For instance, the machine agreed with researchers
about 75% of the time about the difference between light sleep, REM
sleep, and deep sleep. This looks pretty good when you learn that the
two sleep researchers agreed with each other 85% of the time.

[Update, December 11, 2009: Rubin, in a comment, has posted the

research studies

from which I took those statistics.]

The goal of marketing the Zeo, of course, is to help people benefit
from their new knowledge. Typically, a customer records sleep for a
week, along with information about his lifestyle and related
information (such as whether family members woke him up). Zeo offers
a web site where customers can record this data, which was developed
through partnerships with the world’s leading sleep researchers and
consists of fairly standard questions: Are you stressed? Did you drink
caffeine or alcohol in the evening? How do you feel when you first
wake up?

The web site processes the data and makes recommendations, based on
well-understood findings in the field of sleep research, about what
you can change to improve to sleep. Another program you can sign up
for–Sleep Coach–compares the personal behavior you report with your
sleep patterns and makes more targeted recommendations.

Naturally, technologically sophisticated users want to get their hands
on their own data. Zeo has responded by offering customers
data–starting yesterday–in a CSV format. And there’s no reason that
people can’t share their data in order to search for new patterns in a
crowdsourcing manner, like the well-known
PatientsLikeMe
site.

Our new sensitivity to health issues is sure to draw in companies with
technological solutions.
WIRED Magazine
profiled the Nike+
sensor and Internet recording system, and a company named
Withings
recently announced a WiFi-primed body weight scale that records your
weight, body fat content, and BMI to a web site (rather slim benefits
for $160, it seems to me).
Someday we may even be able to monitor many of the medical signs that
currently require blood work through

a device placed on the eye
.

These, along with Zeo, are examples where instant feedback and the
ability to track trends over time can change our whole attitude toward
our bodies. Fitness is no fad; we are recognizing the serious effects
lifestyle can have on our long-term happiness and ultimately our
economic output. Going deeper, it’s about time that computer
technology, which for so long has kept us immobile in our chairs and
divorce us from the physical world, made a 180° turn to put us
more in touch with our bodies and the reality around us.

So far as the Zeo goes, I pondered the value of the data it offers
when you’re considering life style choices. Do you really need a
machine to push you into giving up coffee or turning off the TV
earlier in the evening?

I think data can help, because different people are affected by
lifestyle decisions in different ways. There’s no reason to kick the
dog out of your bedroom or give up that cherished glass of Chianti you
have with your lasagna dinner if it really isn’t bothering you. That’s
the value of feedback: we can link our behavior to an outcome and make
rational choices. It’s easier to act healthy if you know you’re not
just being superstitious.

tags: , , , , , ,
  • http://www.serpentine.com/blog Bryan O'Sullivan

    Harrumph. When I see claims about “complex, proprietary algorithms”, I invariably think “and here’s some some cockamamie handwaving we made up to try to sound impressive”.

  • http://praxagora.com/andyo/ Andy Oram

    Bryan: I don’t usually go for hype either, and I’d certainly prefer to report precisely what they’re doing. But look at it this way: when someone goes to a lab for a traditional sleep test, their face gets covered with all manner of sensors picking up a variety of stimuli. Zeo can’t get all that stuff (which is one reason it’s not used for real medical treatment), but it gets a lot of the same information by unpacking the frequencies from a single wave. Sounds complex to me.

  • jol

    there is a d-quote in the zeo url. interesting read.

    AO: Thanks–fixed now.

  • Luis Romero

    Nice technology, it’s a pity Zeo can’t work as an intelligent awake system (it seems it would be a simple add) I don’t know how much comfortable the headband is (kind of a Heisenberg uncertainty here) but for sure sleep time is quite important and deserves more of those gadgets.

  • http://gradstein.info Lior Gradstein

    Luis,
    Zeo CAN work as an intelligent awake system, look at the bottom of this page:
    http://www.myzeo.com/pages/47_technology.cfm

    “With our optional SmartWake alarm feature, our SoftWave sensors search to find what could be a “natural awakening point” – when it could be a little easier to get out of bed in the morning. SmartWake will look for a moment when you transition into and out of REM sleep when the brain is highly active, a moment Zeo is uniquely able to find since it detects all phases of sleep. The result may be a slightly easier way to wake up. Zeo will awaken you as early as a half-hour before your set wake-up time, and SmartWake will never wake you later than your set wake-up time.”

  • http://CulturalEngineer.blogspot.com Tom Crowl

    Actually looks like an interesting bit of feedback I wouldn’t mind having.

    On the neural network “training” I’d wonder about an inherent cultural bias which I’d hate to see perpetuated…

    Our patterns of awake for 16 hours and then asleep for eight does not actually conform to the way we likely lived and slept as hunter/gatherers.

    Much more typical would be a pattern of intermittent catnaps with, a preponderance certainly at night but typically without anything like the regularity we expect of ourselves.

  • http://darkcoidng.net Graham King

    This article is filled with vague marketing terms: ‘neural network trained on huge numbers of people’ (what is ‘huge’?), algorithms ‘more complex’ than FFT (complex how – Cyclomatic Complexity? Big O notation?), ‘sleep expert’, ‘leading sleep researchers’, etc.

    Is there any published, peer-reviewed research concerning this device? Or are buyers ‘just being superstitious’?

    The general tone of the article is one of product placement, which feels out of place on the O’Reilly Radar.

  • http://www.FloatingBones.com FloatingBones

    This monitor may be effective compared to doing nothing. I would be much more interested in a study comparing it to something like soundersleep.com.

    All things equal, enhancing our natural awareness to monitor and control our bodies is preferable to using bright shiny objects. A skill is preferable to a widget: you never need to replace the batteries or worry about the cat hiding it somewhere. Reliable and repeatable skills to lower stress can be used any time of the day.

    Also, how does it compare to something like the emwave system sold by heartmath.com? Would the ability to monitor your stress instantly from your laptop at any time be more useful than a sleep-monitoring program?

  • Bill Prenovitz

    There’s actually quite a bit of serious research exploring less invasive (and expensive) methods to determine sleep quality. I wonder if Zeo is based on that research or just leveraging the buzz.

  • Ben Rubin

    I just posted a blog article with a breakdown of how Zeo sensors and algorithms work – check it out: http://blog.myzeo.com/5-steps-to-phasing-sleep/

    Regarding the research vs buzz comment – check out our scientific validation section here:
    http://www.myzeo.com/pages/52_for_health_professionals.cfm#Scientific%20Validations

    Best, Ben
    CTO and Co-Founder Zeo Inc.

  • Minnesota Dan

    Can the ZEO headband be used with a CPAP mask?

    Regards,

    Dan

  • Tela

    The ZEO is interesting technology. It seems like you can categorize sleep technology as either passive or active. Zeo, like the Sleep Tracker and other systems are passive but should be part of the kit for someone with serious sleep issues. There aren’t many active sleep systems out there like @FloatingBones suggestion of Sounder Sleep, or the NightWave Sleep Assistant. I think for an active sleep technology to work, it needs to be:
    1. Unobtrusive, and culturally-acceptable
    2. Easy to use
    3. Work by activating the processes the body naturally has for falling asleep.